my p5a-b doesn't work either. It will bsod or freeze randomly with an AGP card. Works fine with PCI cards. I also have the ATX version, and that one works fine.

It might have been overheating... the last time when I shut it down and went to BIOS to check temps it was showing 65C...

Just got another explorer crash after swapping the heatsink. Also PCMark2002 throws errors that it cannot find certain files but installation went perfectly fine... yay for things like that!

I'd open the case up, point a huge fan at it and try to run stuff again. If its overheating it should work fine like that.
Also, if the ATX version isn't as unstable, and it only happens with AGP, it could be that the AT version uses a voltage regulator for the AGP slot that might not be up to the task. I'd try a weaker AGP card (maybe a Banshee? Or a low power Trio3D card?) and see if the issues persist.

I'd advise checking your power supply, or testing it with a known good PSU.

TMPIN0 is the fan.
TMPIN1 is the socket.
TMPIN2 is the CPU core.
Your CPU is rated for up to 85c, so those lower temps shouldn't be a problem.

I recently bought a new in box P5A-B with a K6-III+ 450 ACZ and it just started to throw up crazy errors too.
It got to the point where I couldn't even do a format and re-install of windows 98se.

Run Memtest86.
If you see errors, change out the ram and try again.
If you still get errors, it's bad capacitors. Replace all of them with a known good brand.
The board only uses 1000uf 8.3v 105c rated capacitors, and has 14 of them on the Rev 1.04 board with no onboard audio.

I just replaced my caps with some nice nichicon 1000uf 10v 105c rated ones, and it seems to have cleared up the errors. I was able to run 3 complete passes with memtest and not see a single error.

DNSDies wrote:

TMPIN0 is the fan. TMPIN1 is the socket. TMPIN2 is the CPU core. Your CPU is rated for up to 85c, so those lower temps shouldn't […]
Show full quote

TMPIN0 is the fan.
TMPIN1 is the socket.
TMPIN2 is the CPU core.
Your CPU is rated for up to 85c, so those lower temps shouldn't be a problem.

I recently bought a new in box P5A-B with a K6-III+ 450 ACZ and it just started to throw up crazy errors too.
It got to the point where I couldn't even do a format and re-install of windows 98se.

Run Memtest86.
If you see errors, change out the ram and try again.
If you still get errors, it's bad capacitors. Replace all of them with a known good brand.
The board only uses 1000uf 8.3v 105c rated capacitors, and has 14 of them on the Rev 1.04 board with no onboard audio.

I just replaced my caps with some nice nichicon 1000uf 10v 105c rated ones, and it seems to have cleared up the errors. I was able to run 3 complete passes with memtest and not see a single error.

Thanks for the explanation. I tested RAM and found two sticks with errors on the 1st pass. I'm now using a 256MB stick that didn't throw any errors and we'll see how that goes.

So far Opera 10.10 still crashes a lot so I uninstalled and installed Opera 9.64 which seems to be working fine.

Caps appear OK but I guess replacing them wouldn't hurt. I just recently bought a lot of new caps from Farnell, mostly Rubycon, Panasonic and Nichion. Should have enough for a recap.

Cheers

DNSDies wrote:

The board only uses 1000uf 8.3v 105c rated capacitors, and has 14 of them on the Rev 1.04 board with no onboard audio.

I just replaced my caps with some nice nichicon 1000uf 10v 105c rated ones, and it seems to have cleared up the errors. I was able to run 3 complete passes with memtest and not see a single error.

That's strange, because the board uses Rubycon capacitors which are good quality.

I'd highly recommend trying a good PSU before messing with the caps.

Rubycon or not, they're minimum 15 years old, too, and who knows how many hours of service. I don't think they had
When I tested my caps after removing them, most of them were fine, but there were two near the RAM slots that tested at 900-920uf, just below the 20% margin most caps have. The caps themselves looked fine, too. No bulging or leaking.

It didn't seem very significant, but it was enough to randomly cause memory moves off by 20 bits in the same places every time, regardless of what memory I used. Changing the timings to be slower helped, but it would always bug out before the end of the first pass of memtest.

Now, it works perfectly.

Anyway, if you have a decent setup, the cost of replacing caps is negligible.

DNSDies wrote:

Rubycon or not, they're minimum 15 years old, too, and who knows how many hours of service. I don't think they had
When I tested my caps after removing them, most of them were fine, but there were two near the RAM slots that tested at 900-920uf, just below the 20% margin most caps have. The caps themselves looked fine, too. No bulging or leaking.

That's true, but quality caps from that time do tend to hold up well, even after 15 years. The only way to know for sure would be to test the ESR. Can you please tell me which series of Rubycon they were? My P5A (not P5A-B) has Rubycon YXG.

I would try:

Different psu
Up the voltage 0.1v (might need it for stability)
Lower the voltage 0.1v (might be stable with less voltage and produces less heat)
Try the cpu at 500mhz (some normal k6-2 were not stable at 550mhz, maybe this k6-2+ has the same problem. Should be fixed with better cooling and more voltage)
Try the cpu at a little lower fsb.

If the PSU is fine, the CPU is fine, the ram is fine, your problem is the mainboard. Out of five Ali Aladdin board I own only one works. All the rest have symptoms similar to what you are describing. I've done everything - recapped, changed the VRMs, everything. These things just don't age well. The only Ali board I have that works is an Asus P5A (ATX version). My P5A-B, GA-5AA, S7AX and MS5169 all exhibit various symptoms, from "Gate A20 error when trying to load Himem.sys to BSODs in windows. The only one that "kind of works" out of the four is the MSI, witch runs fine with PCI video cards but will BSOD or lock up in windows with any kind of AGP cards including a Voodoo Banshee. The P5A works as intended, but I wonder how long it will last.

The Gigabyte lasted 2-3 months, then started exhibiting random memory errors. The Commate S7AX came in defective. The P5A-B lasted about two weeks. It failed when I replaced the Voodoo 3 it was running with a Geforce 2 (BSOD after installing forceware drivers - tried to reinstall windows but got errors during installation, then Gate A20 error after CMOS reset, just like the S7AX). The MSI lasted about a months or so.

kanecvr wrote:

The P5A-B lasted about two weeks. It failed when I replaced the Voodoo 3 it was running with a Geforce 2 (BSOD after installing forceware drivers - tried to reinstall windows but got errors during installation, then Gate A20 error after CMOS reset, just like the S7AX).

This one most probably needs a new 3.3v regulator for the AGP power supply needs.
These boards had really weak ones. Does it get any hot when an AGP card is inserted? Do you measure any 3.3v?

Quick update:
after finding out that several RAM sticks were reported as bad by memtest and using one that doesn't give any errors, looks like the system works. I'll leave it at that for now.

Could be timing issues with the RAM due to failing components.
You could try manually setting the timing to a more forgiving speed if the problem shows up again.

If it works, though, just leave it be. Older machines can be temperamental.